1 /*
2  * Copyright (c) 2007-2012 Nicira, Inc.
3  *
4  * This program is free software; you can redistribute it and/or
5  * modify it under the terms of version 2 of the GNU General Public
6  * License as published by the Free Software Foundation.
7  *
8  * This program is distributed in the hope that it will be useful, but
9  * WITHOUT ANY WARRANTY; without even the implied warranty of
10  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11  * General Public License for more details.
12  *
13  * You should have received a copy of the GNU General Public License
14  * along with this program; if not, write to the Free Software
15  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
16  * 02110-1301, USA
17  */
18 
19 #include <linux/if_vlan.h>
20 #include <linux/kernel.h>
21 #include <linux/netdevice.h>
22 #include <linux/etherdevice.h>
23 #include <linux/ethtool.h>
24 #include <linux/skbuff.h>
25 
26 #include <net/dst.h>
27 #include <net/xfrm.h>
28 #include <net/rtnetlink.h>
29 
30 #include "datapath.h"
31 #include "vport-internal_dev.h"
32 #include "vport-netdev.h"
33 
34 struct internal_dev {
35 	struct vport *vport;
36 };
37 
38 static struct vport_ops ovs_internal_vport_ops;
39 
40 static struct internal_dev *internal_dev_priv(struct net_device *netdev)
41 {
42 	return netdev_priv(netdev);
43 }
44 
45 /* Called with rcu_read_lock_bh. */
46 static netdev_tx_t
47 internal_dev_xmit(struct sk_buff *skb, struct net_device *netdev)
48 {
49 	int len, err;
50 
51 	len = skb->len;
52 	rcu_read_lock();
53 	err = ovs_vport_receive(internal_dev_priv(netdev)->vport, skb, NULL);
54 	rcu_read_unlock();
55 
56 	if (likely(!err)) {
57 		struct pcpu_sw_netstats *tstats = this_cpu_ptr(netdev->tstats);
58 
59 		u64_stats_update_begin(&tstats->syncp);
60 		tstats->tx_bytes += len;
61 		tstats->tx_packets++;
62 		u64_stats_update_end(&tstats->syncp);
63 	} else {
64 		netdev->stats.tx_errors++;
65 	}
66 	return NETDEV_TX_OK;
67 }
68 
69 static int internal_dev_open(struct net_device *netdev)
70 {
71 	netif_start_queue(netdev);
72 	return 0;
73 }
74 
75 static int internal_dev_stop(struct net_device *netdev)
76 {
77 	netif_stop_queue(netdev);
78 	return 0;
79 }
80 
81 static void internal_dev_getinfo(struct net_device *netdev,
82 				 struct ethtool_drvinfo *info)
83 {
84 	strlcpy(info->driver, "openvswitch", sizeof(info->driver));
85 }
86 
87 static const struct ethtool_ops internal_dev_ethtool_ops = {
88 	.get_drvinfo	= internal_dev_getinfo,
89 	.get_link	= ethtool_op_get_link,
90 };
91 
92 static void internal_dev_destructor(struct net_device *dev)
93 {
94 	struct vport *vport = ovs_internal_dev_get_vport(dev);
95 
96 	ovs_vport_free(vport);
97 }
98 
99 static void
100 internal_get_stats(struct net_device *dev, struct rtnl_link_stats64 *stats)
101 {
102 	int i;
103 
104 	memset(stats, 0, sizeof(*stats));
105 	stats->rx_errors  = dev->stats.rx_errors;
106 	stats->tx_errors  = dev->stats.tx_errors;
107 	stats->tx_dropped = dev->stats.tx_dropped;
108 	stats->rx_dropped = dev->stats.rx_dropped;
109 
110 	for_each_possible_cpu(i) {
111 		const struct pcpu_sw_netstats *percpu_stats;
112 		struct pcpu_sw_netstats local_stats;
113 		unsigned int start;
114 
115 		percpu_stats = per_cpu_ptr(dev->tstats, i);
116 
117 		do {
118 			start = u64_stats_fetch_begin_irq(&percpu_stats->syncp);
119 			local_stats = *percpu_stats;
120 		} while (u64_stats_fetch_retry_irq(&percpu_stats->syncp, start));
121 
122 		stats->rx_bytes         += local_stats.rx_bytes;
123 		stats->rx_packets       += local_stats.rx_packets;
124 		stats->tx_bytes         += local_stats.tx_bytes;
125 		stats->tx_packets       += local_stats.tx_packets;
126 	}
127 }
128 
129 static const struct net_device_ops internal_dev_netdev_ops = {
130 	.ndo_open = internal_dev_open,
131 	.ndo_stop = internal_dev_stop,
132 	.ndo_start_xmit = internal_dev_xmit,
133 	.ndo_set_mac_address = eth_mac_addr,
134 	.ndo_get_stats64 = internal_get_stats,
135 };
136 
137 static struct rtnl_link_ops internal_dev_link_ops __read_mostly = {
138 	.kind = "openvswitch",
139 };
140 
141 static void do_setup(struct net_device *netdev)
142 {
143 	ether_setup(netdev);
144 
145 	netdev->max_mtu = ETH_MAX_MTU;
146 
147 	netdev->netdev_ops = &internal_dev_netdev_ops;
148 
149 	netdev->priv_flags &= ~IFF_TX_SKB_SHARING;
150 	netdev->priv_flags |= IFF_LIVE_ADDR_CHANGE | IFF_OPENVSWITCH |
151 			      IFF_NO_QUEUE;
152 	netdev->needs_free_netdev = true;
153 	netdev->priv_destructor = internal_dev_destructor;
154 	netdev->ethtool_ops = &internal_dev_ethtool_ops;
155 	netdev->rtnl_link_ops = &internal_dev_link_ops;
156 
157 	netdev->features = NETIF_F_LLTX | NETIF_F_SG | NETIF_F_FRAGLIST |
158 			   NETIF_F_HIGHDMA | NETIF_F_HW_CSUM |
159 			   NETIF_F_GSO_SOFTWARE | NETIF_F_GSO_ENCAP_ALL;
160 
161 	netdev->vlan_features = netdev->features;
162 	netdev->hw_enc_features = netdev->features;
163 	netdev->features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
164 	netdev->hw_features = netdev->features & ~NETIF_F_LLTX;
165 
166 	eth_hw_addr_random(netdev);
167 }
168 
169 static struct vport *internal_dev_create(const struct vport_parms *parms)
170 {
171 	struct vport *vport;
172 	struct internal_dev *internal_dev;
173 	int err;
174 
175 	vport = ovs_vport_alloc(0, &ovs_internal_vport_ops, parms);
176 	if (IS_ERR(vport)) {
177 		err = PTR_ERR(vport);
178 		goto error;
179 	}
180 
181 	vport->dev = alloc_netdev(sizeof(struct internal_dev),
182 				  parms->name, NET_NAME_USER, do_setup);
183 	if (!vport->dev) {
184 		err = -ENOMEM;
185 		goto error_free_vport;
186 	}
187 	vport->dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
188 	if (!vport->dev->tstats) {
189 		err = -ENOMEM;
190 		goto error_free_netdev;
191 	}
192 
193 	dev_net_set(vport->dev, ovs_dp_get_net(vport->dp));
194 	internal_dev = internal_dev_priv(vport->dev);
195 	internal_dev->vport = vport;
196 
197 	/* Restrict bridge port to current netns. */
198 	if (vport->port_no == OVSP_LOCAL)
199 		vport->dev->features |= NETIF_F_NETNS_LOCAL;
200 
201 	rtnl_lock();
202 	err = register_netdevice(vport->dev);
203 	if (err)
204 		goto error_unlock;
205 
206 	dev_set_promiscuity(vport->dev, 1);
207 	rtnl_unlock();
208 	netif_start_queue(vport->dev);
209 
210 	return vport;
211 
212 error_unlock:
213 	rtnl_unlock();
214 	free_percpu(vport->dev->tstats);
215 error_free_netdev:
216 	free_netdev(vport->dev);
217 error_free_vport:
218 	ovs_vport_free(vport);
219 error:
220 	return ERR_PTR(err);
221 }
222 
223 static void internal_dev_destroy(struct vport *vport)
224 {
225 	netif_stop_queue(vport->dev);
226 	rtnl_lock();
227 	dev_set_promiscuity(vport->dev, -1);
228 
229 	/* unregister_netdevice() waits for an RCU grace period. */
230 	unregister_netdevice(vport->dev);
231 	free_percpu(vport->dev->tstats);
232 	rtnl_unlock();
233 }
234 
235 static netdev_tx_t internal_dev_recv(struct sk_buff *skb)
236 {
237 	struct net_device *netdev = skb->dev;
238 	struct pcpu_sw_netstats *stats;
239 
240 	if (unlikely(!(netdev->flags & IFF_UP))) {
241 		kfree_skb(skb);
242 		netdev->stats.rx_dropped++;
243 		return NETDEV_TX_OK;
244 	}
245 
246 	skb_dst_drop(skb);
247 	nf_reset(skb);
248 	secpath_reset(skb);
249 
250 	skb->pkt_type = PACKET_HOST;
251 	skb->protocol = eth_type_trans(skb, netdev);
252 	skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
253 
254 	stats = this_cpu_ptr(netdev->tstats);
255 	u64_stats_update_begin(&stats->syncp);
256 	stats->rx_packets++;
257 	stats->rx_bytes += skb->len;
258 	u64_stats_update_end(&stats->syncp);
259 
260 	netif_rx(skb);
261 	return NETDEV_TX_OK;
262 }
263 
264 static struct vport_ops ovs_internal_vport_ops = {
265 	.type		= OVS_VPORT_TYPE_INTERNAL,
266 	.create		= internal_dev_create,
267 	.destroy	= internal_dev_destroy,
268 	.send		= internal_dev_recv,
269 };
270 
271 int ovs_is_internal_dev(const struct net_device *netdev)
272 {
273 	return netdev->netdev_ops == &internal_dev_netdev_ops;
274 }
275 
276 struct vport *ovs_internal_dev_get_vport(struct net_device *netdev)
277 {
278 	if (!ovs_is_internal_dev(netdev))
279 		return NULL;
280 
281 	return internal_dev_priv(netdev)->vport;
282 }
283 
284 int ovs_internal_dev_rtnl_link_register(void)
285 {
286 	int err;
287 
288 	err = rtnl_link_register(&internal_dev_link_ops);
289 	if (err < 0)
290 		return err;
291 
292 	err = ovs_vport_ops_register(&ovs_internal_vport_ops);
293 	if (err < 0)
294 		rtnl_link_unregister(&internal_dev_link_ops);
295 
296 	return err;
297 }
298 
299 void ovs_internal_dev_rtnl_link_unregister(void)
300 {
301 	ovs_vport_ops_unregister(&ovs_internal_vport_ops);
302 	rtnl_link_unregister(&internal_dev_link_ops);
303 }
304